#ai

GPT-5.5 landed a week after Claude Opus 4.7, both claiming top of leaderboard. The category scores tell a cleaner story. The two models split coding work down the middle, and the split holds across independent harnesses. Pick by task, not by top-line score.

Apple named John Ternus CEO on April 20, 2026. The press called it smooth succession. The sharper read is that Apple already conceded its AI strategy in January with a $1 billion a year Gemini deal, and Ternus is the CEO you pick to ship that concession cleanly.

UT Austin just pledged $750 million for what it's calling the first 'AI-native' hospital, opening 2030. The press release says almost nothing about what that means. Trade-press coverage the same day named five specific systems. Whether the bet is real turns on the gap between them.

MIT's Recursive Language Models load the prompt into a Python REPL the model queries with code. RLM(GPT-5-mini) beats GPT-5 on long-context benchmarks at similar cost. The argument: the context-window arms race has been solving the wrong problem.

Anthropic's new Claude Design reads your codebase on the way in and exports to Canva, Claude Code, and PPTX on the way out. Figma (FIG) fell 6.89% the day it launched. The way the product is scoped tells you what Anthropic thinks AI can and can't own in design.

Anthropic shipped Claude Code routines on April 14, 2026: a research-preview feature that turns any prompt into a cloud-hosted agent triggered by schedule, HTTP POST, or GitHub event. Here is what it is, how to set one up, and where the daily caps actually bite.

Beyond basic prompting: system prompt architecture, structured outputs, RAG patterns, LLM-as-judge evaluation, and cost optimization for production AI.

A rigorous look at the semiconductor investment landscape: NVIDIA's moat, the TSMC bottleneck, HBM dynamics, and how to value chip companies in an AI cycle.

AI underwriting, computer vision for property assessment, NLP lease review, and predictive maintenance are moving from pilot to production. Here's what's working and what's still overhyped.

AI coding tools haven't eliminated engineering headcount, but they've changed what good looks like. Here's how to hire, retain, and level engineers in a world where the tools do more.

GPUs get the headlines, but inference serving, networking, and observability are where AI systems actually win or lose. Here's what the stack looks like now.